The pay cut problem
Your law firm is not slow to go AI-native because it lacks talent. It is slow because going AI-native requires the partners to vote for a smaller cheque.
Jake Saper at Emergence Capital put the sharpest version of this in one line: the reason your law firm will not go AI-native is not talent, it is that the partners would have to take a pay cut. (his post)
It is worth sitting with, because it explains a pattern that otherwise looks irrational. These firms employ some of the most capable people in the economy. They are not confused about what AI does. Many have run pilots, hired heads of innovation, and published thoughtful memos about the future of the profession. And the delivery model at the bottom has barely moved.
That is not a competence failure. It is arithmetic.
Where the money goes
A partnership is designed to distribute its profits. That is the point of the structure. Each year the surplus is calculated and paid out to the equity partners. There is no large retained pool, because retaining earnings is the thing a partnership specifically does not do.
So when you ask a partnership to fund a multi-year rebuild of how the work gets done, you are asking a specific group of people to vote to pay themselves less, for several years, in order to hand a more efficient firm to whoever holds equity later. Many of the partners voting are within a decade of drawing their last distribution. The costs land on them and the benefits land on their successors.
Now add the second problem, which is worse. The billable hour means the firm's revenue is a function of time spent. Any technology that cuts the hours needed cuts the revenue attached to them, unless the firm simultaneously repositions how it charges. So the successful pilot arrives on the management committee's desk carrying a genuinely awkward sentence: this works, and it will shrink the top line.
Set against that, "let us go carefully and monitor developments" is not cowardice. It is the rational move for the people in the room.
My read, not a measured claim: this is why the visible AI activity at incumbent firms clusters in research, drafting assistance, and knowledge search, and thins out sharply wherever it would touch how the work is priced. Those are the places where efficiency does not threaten the revenue model.
Why the challenger has it easier
A firm built around the system from the start has none of those constraints. There is no partner distribution to protect, because there are no partners drawing one. There is no hour to defend, because it never sold hours. It can price against the outcome from day one, which is only possible if the system measures what it recommends, which is only possible if you built it that way at the beginning.
None of that is cleverness. It is the absence of a constraint the incumbent cannot remove without a vote it will not win.
And it is the reason the useful question is not whether traditional firms understand AI. They do. The useful question is who is structurally able to act on it.
What this means if you are the buyer
You are not choosing between a firm that gets AI and a firm that does not. You are choosing between two economic models, and the model determines what you can ask for.
Three questions separate them quickly.
What am I actually paying for? Hours, or a result. If the answer is hours, efficiency is not on your side of the table.
What happens when the advice is wrong? In a time-billing model, usually nothing. The invoice already cleared. Ask whether there is a record, whether predictions are written down before the fact, and whether anyone goes back to check.
Who does the work, and what happens when they leave? If the answer is a team of juniors and a partner's judgment, then the firm's memory of your business is a person, and it walks when they do.
None of these are gotchas. A good traditional firm will answer them honestly, and for some problems it remains the right choice. But the answers tell you which model you are buying, and the model is what you are stuck with.
The part nobody says out loud
The incumbents are not going away. Partnerships are resilient, relationships are real, and there is work where a named human with thirty years of judgment is worth every pound.
What is changing is the range of work where that is true. It is narrowing, and it is narrowing from the bottom, in the layer that used to be juniors reading, summarising, and drafting at volume. That layer was also where the margin came from.
Which is the actual thing to watch. Not whether the big firms adopt AI. Whether they can reprice before the layer they earn on is gone.